33,347 karma · joined November 1, 2011
We’re not lacking for opportunities to believe the worst of each other. It’s not something TikTok invented.
Most people attacking the Capitol were unorganized protestors, who were gathered as cover for the real effort, which was quite well organized. It was well described in Congressional hearings; people were convicted of conspiracy; lawyers have been disbarred, etc.
So that pretty much kills any explanation that depends on our recent experience, like pressure to get a degree, good job, etc. Seems like there is probably something far more fundamental at work to create a 200+ year trend with a one brief interruption for a couple decades.
My point is that it’s very very hard to know in advance that a startup is going to be awesome at marketing. Their tech was not, on its own, a giant moat or a “must have” standout. So I can see why some VCs shrugged.
My impression was that LLM training codebases were 99% resource management and only a few lines actually implement the core training algorithm, which is where 100% of the intelligence comes from. Data, not lines of code, are the constraint.
After training you can adapt the intelligence in various ways, and that takes a bunch of lines of coded too. But you cant raise the intelligence ceiling again without another training run. So where is the scary recursive part?
Companies do tons of communication and work directly, without press releases or blog posts. If a statement is released publicly, it is done for a PR purpose.
Note that the episode is NOT about the ship’s computer. They all know it’s not conscious, despite also being a machine that can converse and do things.
Our LLMs are like the ship’s computer.
They make not-crappy productivity tools at not-cheap price points, and aim for top-5 market share. That’s not a luxury product strategy. They are a lot more like Honda or Volkswagon than Lamborghini.
I think if you look through this thread you’ll see a lot of skepticism of the AI results, and I think that is a fairly broadly held opinion. The obvious way to check the AI answer is to click through to some sources.
I think for Google to stop sending me traffic, it would have to be essentially perfect at AI answers. It will never get there, especially as so many searches are opinion-based like “what is the best mobile phone right now.”
My go-to analogy is the concept of heat and warmth, which was the subject of philosophical speculation and inquiry for centuries. But ultimately it was not useful or predictive until it was redefined very narrowly by scientists as a specific measurable attribute of a collection of molecules. Arguably this redefinition itself was an important component of the establishment of “science” as a discipline distinct from philosophy.
This let scientists work with heat rigorously, even though non-scientists still have a largely colloquial vocabulary about it, e.g. “this coat is warmer than my last one.” On matters like consciousness, we’re all still smashed together into colloquial land, arguing about definitions when we think we are arguing about concepts.
I do think the current advancements in AI may help us develop the new vocabulary we need to quantitatively reason about intelligence and language, and in doing so, help better define / constrain what the we mean by consciousness. We think of LLMs now as a tool we can use to do work, but another view is that they are a sensor that directly correlates physical inputs to intelligent outputs. Sort of like how burning can be used to measure the energy contained in various materials, in addition to keeping us warm.
In science for example, anyone can do an experiment about gravity. In fact millions of high school kids do every year. What makes an expert scientist is the ability to understand all the many ways such experiments can fail to accurately measure the underlying reality.
Or consider an AI writing a press release. A PR expert will catch nuances of wording that will confuse readers, or leave fodder for others to attack or mischaracterize the announcement.
College students know this because they are working with AI. And what makes them mad is the human-driven false notion that AI devalues expertise. AI looks like magic to non-experts. But it’s not, it’s more like a “junior engineer” or “PR intern” to people with actual expertise to evaluate its output.
We configured a new email sending service and kept the DNS TTLs low on the TXT records for SPF, DKIM, and DMARC, in case we needed to change them. We saw a lot of mystery failures for emails going to Microsoft inboxes (M365 and Outlook.com). Changing the TTLs to be very long (86400 or more) caused a large improvement within a day to two.
The only way I can think to explain this is that some of their DNS lookups would time out if they had to follow recursion back to our DNS provider. Lengthening the TTL increased the chance the records would be cached locally to Microsoft’s systems and therefore served faster.
The only other explanation I can think of is that MS prefers longer TTLs as a matter of policy and downgrades based on that. But usually they publish policy preferences like that and I could not find one.
Unfortunately, a lot of people still believe in widespread voter fraud despite the obvious and well-documented rarity.
I would say something similar about the availability of voter files, though. A lot of people are horrified they are available despite the obvious and well-documented lack of evidence it causes any significant harm.
This sort of determinism has been a problem for Dawkins going back to the late chapters of The Selfish Gene.
It took something far more powerful than a TV show that was mildly popular for a few years to create what we live today. It took the great flatness of the Internet. As Terry Pratchett predicted to Bill Gates back then.
This essentially is where the schism is between science and philosophy, and has played out repeatedly across history. Heat for example was redefined to a specific physical property, and subjective experiences of warmth were then explored in reference to that. Or look back to the moment when Newton essentially said “I don’t know what gravity is, but I can accurately calculate any ballistic trajectory you can think of.”
That said, pretty much everything about highest-end athletics is net negative for long-term health. It’s incredibly hard on the body to run a marathon in general, let alone at record breaking pace.
It’s not that different from the general concept of pricing. People will swear they want to buy American, support small business etc. but when it’s time for new jeans they go to Walmart and buy the pair on sale.
The belief is that the first country to have this reliably at scale breaks the “mutually assured destruction” paradigm that has governed nuclear weapons policy for decades. If the U.S. can send nuclear ballistic missiles, but can’t be hit by nuclear missiles, what stops them from just nuking anyone who disagrees with them?